[14p-P10-1~25] Symposium on Research and Human-Capital Development Activities aimed at Creating Next-Generation Semiconductors Industry

English Presentation

[14p-P10-14] Accelerating DNN Models with DPU: A Hardware-Software Co-Design Approach

〇JIAWEI Yu¹, Atsutake Kosuge¹, Hunga Amano¹ (1.Kosuge Lab)

Keywords:

Edge device

With the growth of IoT applications, there is an increasing need for real-time object detection on edge devices. Edge computing has emerged as a solution to the traditional cloud computing challenges of high latency, limited bandwidth, and privacy risks by moving computation to end devices. Among edge computing platforms, FPGA-based DPU solutions stand out as an ideal choice for deploying deep learning models, thanks to their low power consumption, high performance, and reconfigurability.